Perhitungan Kendaraan Menggunakan Yolo Dari Ultralytics Untuk Sistem Pemantauan Lalu Lintas
Keywords:
Class Imbalance, Deteksi Kendaraan, Intelligent Transportation System, Random Oversampling, YOLO26s.Abstract
Pertumbuhan volume kendaraan di perkotaan memerlukan sistem pemantauan otomatis seperti Intelligent Transportation System (ITS). Namun, model deep learning sering menghadapi kendala ketidakseimbangan kelas (class imbalance) ekstrem yang menurunkan akurasi objek minoritas. Penelitian ini menerapkan arsitektur ringan YOLO26s dengan strategi random oversampling untuk mendeteksi empat kelas kendaraan: Bus, Mobil, Sepeda Motor, dan Truk. Dataset gabungan sejumlah 19.389 citra bersumber dari UA-DETRAC dan Roboflow, lalu dilatih memakai konfigurasi Distributed Data Parallel (DDP) pada dua GPU NVIDIA Tesla T4. Hasil pengujian membuktikan teknik oversampling meningkatkan performa model secara signifikan tanpa memicu overfitting. Nilai mAP@0.5 global meningkat dari 0,982 menjadi 0,993, dan mAP@0.5:0.95 naik dari 0,842 menjadi 0,911. Peningkatan krusial terjadi pada Recall kelas Truk yang melonjak dari 0,93 menjadi 0,98. Model juga mempertahankan efisiensi tinggi dengan kecepatan inferensi mencapai 85,51 FPS (11,69 ms), sehingga sangat ideal untuk implementasi pemantauan lalu lintas secara real-time.
References
[1] G. Jocher, A. Chaurasia, and J. Qiu, "Ultralytics YOLO," version 8.0, 2023. [Online]. Available: https://github.com/ultralytics/ultralytics K. Jordan, "Muon: An optimizer for filters," GitHub Repository, 2024. [Online]. Available: https://github.com/kellerjordan/muon C. H. Lim, T. Connie, T. S. Ong, et al., "Visual-Based Vehicle Detection with Adaptive Oversampling," International Journal of Information Technology, vol. 16, pp. 4767–4777, 2024, doi: 10.1007/s41870-024-01977-w. T. Diwan, G. Anirudh, and J. V. Tembhurne, "Object detection using YOLO: Challenges, architectural successors, datasets and applications," Multimedia Tools and Applications, vol. 82, no. 6, pp. 9243–9275, 2023, doi: 10.1007/s11042-022-13644-y. J. Liu, Y. Xie, Y. Zhang, and H. Li, "Vehicle Flow Detection and Tracking Based on an Improved YOLOv8n and ByteTrack Framework," World Electric Vehicle Journal, vol. 16, no. 1, p. 13, 2025, doi: 10.3390/wevj16010013. J. Zhao, J. Hao, X. Zhou, S. Xiao, and B. Wang, "Improved Vision-Based Vehicle Detection and Classification by Optimized YOLOv4," IEEE Access, vol. 10, pp. 8590–8603, 2022, doi: 10.1109/ACCESS.2022.3143365. C. Li dkk., "YOLOv6: A Single-Stage Object Detection Framework for Industrial Applications," arXiv preprint arXiv:2209.02976, 2022. T. Kumar, R. Brennan, A. Mileo, and M. Bendechache, "Image Data Augmentation Approaches: A Comprehensive Survey and Future Directions," IEEE Access, vol. 12, pp. 187536–187571, 2024, doi: 10.1109/ACCESS.2024.3470122. L. Wen, D. Du, Z. Cai, Z. Lei, M.-C. Chang, H. Qi, J. Lim, M.-H. Yang, and S. Lyu, "UA-DETRAC: A New Benchmark and Protocol for Multi-Object Detection and Tracking," Computer Vision and Image Understanding, vol. 193, p. 102907, Aug. 2020, doi: 10.1016/j.cviu.2020.102907. Y. Zhang, P. Sun, Y. Jiang, D. Yu, F. Weng, Z. Yuan, P. Luo, W. Liu, and X. Wang, "ByteTrack: Multi-Object Tracking by Associating Every Detection Box," Proceedings of the European Conference on Computer Vision (ECCV), pp. 1–21, Oct. 2022, doi: 10.1007/978-3-031-20047-2_1.
